(62j) Powering Next-Generation Batteries through Data Science in Synergy with Simulations and Experiments

In response to the escalating demands for advanced energy storage solutions, my research primarily concentrates on pioneering electrolyte discovery for next-generation batteries through the innovative application of artificial intelligence (AI). Leveraging state-of-the-art machine learning (ML) techniques, including shallow learning and deep learning algorithms such as graph neural networks (GNNs), we delve into largest datasets for liquid electrolytes curated in-house from literature. For electrolytes, all three crucial figures of merits were employed for developing separate ML models – ionic conductivity, oxidative stability, and Coulombic efficiency. Due to the disparate nature of optimizing three properties simultaneously, we devised a new metric called the eScore. The eScores were used to find few promising candidates from trained generalizable ML models predictions on vast unlabeled sets of electrolyte molecules. This approach allows us to transcend traditional boundaries and identify novel, high-potential electrolyte candidates within a largely unexplored chemical space. Our AI-driven methodology embodies a transformative shift in battery materials research, moving from the slow, iterative trial-and-error methods of the past to a rapid, data-informed discovery process.